Abstract
Traditional assessments of high-speed craft (HSC) hydrodynamics rely on experimental, semiempirical, and numerical methods, which can be computationally expensive and poorly suited for real-time applications. This paper introduces a Neural Ordinary Differential Equation model for predicting the dynamic response of the Generic Prismatic Planing Hull. Instead of relying on fixed difference equations, the neural ordinary differential equation model learns a continuoustime representation of vessel dynamics directly from experimental data. The model achieves high accuracy in one-step-ahead predictions when provided with ground truth inputs, successfully capturing vessel behavior in both regular wave and calm water conditions. A noiseinjected training scheme was introduced to improve robustness to small input perturbations, enabling consistent performance across a range of sea states. These results highlight the potential of Neural ODEs for real-time simulation and maritime autonomy.